The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill instructs the agent to install the sports-skills Python package from PyPI or directly from the vendor's GitHub repository (github.com/machina-sports/sports-skills.git). These are identified as vendor-owned resources supporting the skill's primary functionality.
[COMMAND_EXECUTION]: The agent uses shell commands to verify installation (which sports-skills), perform package installation (pip install), and execute data queries via a CLI tool.
[DATA_EXFILTRATION]: No evidence of data exfiltration was found. The skill operates on public sports data and does not require or access sensitive user credentials or environment variables.
[PROMPT_INJECTION]: The skill possesses an indirect prompt injection surface due to its ingestion of external data (news articles and play-by-play descriptions) which could theoretically contain malicious instructions.
Ingestion points: Data enters the context via the get_news and get_play_by_play endpoints from ESPN.
Boundary markers: The instructions do not define specific delimiters to isolate external text from the agent's core logic.
Capability inventory: The skill environment allows for command execution and package management.
Sanitization: There is no mention of sanitizing or filtering the content returned from the external API before the agent processes it.

nfl-data